The US blacklisted Anthropic as a security threat. Its spy agencies are using Claude anyway.

The Pentagon has officially blacklisted Anthropic as a national security supply chain threat due to concerns about its corporate structure and foreign investment ties. Despite this, the NSA continues using Claude because a critical shortage of advanced chips prevents spy agencies from running alternative frontier AI models on classified networks. The White House approved a secret $9 billion emergency funding request to build specialized federal data centres powered by Nvidia's Grace Blackwell superchips, with $800 million redirected immediately from other budgets. The contradiction highlights a structural tension in US AI policy: frontier AI capability is concentrated in a few private companies, and when hardware to run alternatives doesn't exist, the government loses leverage over its own supply chain. Anthropic's revenue surged from $9B to $30B annualised in early 2026, and the company is preparing for a potential IPO at an $800B valuation.